Published: December 21, 2010

r 2010 American Chemical Society 1188 dx.doi.org/10.1021/ja108626w | J. Am. Chem. Soc. 2011, 133, 1188–1191

COMMUNICATION

pubs.acs.org/JACS

Global Structure of Forked DNA in Solution Revealed by High-Resolution Single-Molecule FRETTara Sabir,†,‡ Gunnar F. Schr€oder,§ Anita Toulmin,†,‡ Peter McGlynn, ) and Steven W. Magennis*,†,‡

†School of Chemistry, The University of Manchester, Oxford Road, Manchester M13 9PL, U.K.‡The Photon Science Institute, The University of Manchester, Alan Turing Building, Oxford Road, Manchester M13 9PL, U.K.§Institute of Structural Biology and Biophysics, Forschungszentrum J€ulich, 52425 J€ulich, Germany

)Institute of Medical Sciences, University of Aberdeen, Ashgrove Road West, Aberdeen AB25 2ZD, U.K.

bS Supporting Information

ABSTRACT: Branched DNA structures play critical roles inDNA replication, repair, and recombination in addition tobeing key building blocks for DNA nanotechnology. Here wecombine single-molecule multiparameter fluorescence detec-tion and molecular dynamics simulations to give a generalapproach to global structure determination of branched DNAin solution. We reveal an open, planar structure of a forkedDNA molecule with three duplex arms and demonstrate anion-induced conformational change. This structure will serveas a benchmark for DNA-protein interaction studies.

Branched DNA molecules are central intermediates in gen-ome duplication, where a parental duplex is copied to form

two daughter duplexes. They are also central to enzymatic pro-cessing of damaged replication forks, repair of DNA damage, andrecombination between homologous duplexes.1 The structures ofbranched DNA intermediates are therefore likely to impact directlyon recognitionby amultitudeofDNA-processing enzymes. BranchedDNA is also widely used as a building block in DNA nano-structures2 and nanomechanical devices3 as well as in DNAcomputation.4 Although X-ray crystal structures are availablefor a small number of such molecules, notably the Hollidayjunction,5 the detailed structures of branched nucleic acids arelargely unknown.6 We show here that multiparameter fluores-cence detection (MFD) of single molecules in combination withmolecular dynamics (MD) simulations can reveal the accuratethree-dimensional global structure of a three-armed forked DNAmolecule in solution.

In view of the potential for static or dynamic heterogeneity insolution,7 it is desirable to elucidate the structure of individualbranched DNA molecules without ensemble averaging. Atomicforce microscopy (AFM) is very effective at probing single mole-cules in solution and has been successfully used to study DNAnanostructures.4 However, distortion of the structure due tointeraction with the surface cannot be ruled out, particularly forflexible molecules.8 Single-molecule fluorescence resonance en-ergy transfer (SM-FRET), which probes inter- or intramolecularenergy transfer between chromophoric labels on the 1-10 nmlength scale,9 can also be used to study the structure anddynamics of DNA structures that are either immobilized or free

in solution.10 In spite of this, the use of FRET as a quantitativestructural tool is hampered by uncertainties in the positions andorientations of the fluorescent dyes as well as by calibrationfactors such as detection efficiencies and spectral cross-talk.11

The MFD approach measures the color, lifetime, polarization,and intensity of fluorescence simultaneously12 and, together withdetailed MD simulations of local dye positions, has allowedaccurate distance determinations within unbranched double-stranded DNA (dsDNA).13

We used MFD of SM-FRET to study a branched DNAmolecule, which we term a four-stranded fork (4SF), designedto mimic the possible structure of a blocked replication fork.1

The DNA sequence of the 4SF and the positions of thefluorescent dyes are shown in Figure 1a. A geometric model ofthe 4SF is shown in Figure 1b.We considered the branchedDNAas three independent helical arms connected by a freely rotatingjoint, the branch point (see Figure 1b). Orientation of the axisframe of one helical arm with respect to a reference frame ofanother helical arm requires two polar angles and an angle ofrotation around the helical axis; since there are two independenthelical axis frames, we required six independent parameters todefine the global structure.

We studied fully complementary, immobile branched DNAstructures by annealing of single strands that were designed tominimize undesirable base pairing (e.g., hairpin formation) andsequence-specific bending (e.g., A-tracts). For FRET measure-ments, we investigated branched molecules having a single donordye, Alexa488, in one of three positions (D1, D2, or D3) and anacceptor dye, Cy5, in one of three positions (A1, A2, or A3). Thesix dye positions allowed us to measure eight unique distances.We also studied the corresponding donor-only structures tocheck for local quenching effects.

Previous studies of Holliday junctions have revealed a numberof characteristic features of nucleic acid junctions, including theirability to undergo pairwise coaxial stacking and ion-induced folding.6

Holliday junctions can undergo both specific and nonspecificinteractions withmonovalent andmultivalent cations, dependingon the local ion concentration.6 Therefore, we studied the 4SF intwo buffers, one containing no Mg2þ ions and one with 1 mMMgCl2 present.

Received: September 24, 2010

1189 dx.doi.org/10.1021/ja108626w |J. Am. Chem. Soc. 2011, 133, 1188–1191

Journal of the American Chemical Society COMMUNICATION

Figure 2 shows typical MFD data for a donor-only control anda donor-acceptor sample. The MFD method uses a confocalmicroscope with pulsed laser excitation and simultaneous photon-counting detection of fluorescence in four channels (two colorsfor both parallel and perpendicular polarization).12 See theSupporting Information for full details of the experimental setup.The two-dimensional (2D) frequency plots are of FRET effi-ciency (E) or donor anisotropy (rD) versus donor lifetime(τD(A)). MFD allows FRET-related species to be unambiguouslyassigned, since these fall on the theoretical curves highlighted inFigure 2. Photobleaching, dye quenching, and the presence ofimpurities are easily discerned and do not affect the distancemeasurement. For every donor-only sample, we observed onemain population (Figure 2a). As expected, the Alexa488 had alifetime of 4.1 ns at all labeling positions. We also observed aminor population with a lifetime of ∼2 ns, which was observedpreviously in MFD of Alexa488 and is attributed to a quenchedform of the dye.14 The anisotropy of the donor was low, asobserved previously, ruling out dye orientation effects.13 For donor-acceptor samples, two populations were observed (Figure 2b); onewas identical to that of the donor-only sample (due to unlabeled orphotobleached acceptor strand), while the other had a shorterdonor lifetime than the donor-only species as well as a reduceddonor/acceptor intensity ratio and increased anisotropy. TheMFDdata demonstrate unambiguously that this second population wasdue to doubly labeled branched DNA molecules undergoingFRET. The widths of the FRET peaks were similar to those ofshot-noise-limited FRET populations measured previouslyfor dsDNA, suggesting either the absence of a static distancedistribution or fast dynamics.14,15

To convert these accurate FRET measurements into absolutedye-dye distances, we used a nonlinear conversion function that

takes into account the positional and orientational averaging ofthe dyes (see the Supporting Information).13 The calculateddistances are given in Table 1.

These dye-dye distances were used as restraints in MDsimulations. As the deviations due to sequence-dependent

Figure 2. Single-moleculemultiparameter fluorescence detection (MFD)of branched DNA. Typical data are shown for a four-stranded fork labeledwith (a) a donor dye (at position D2 in Figure 1a) and (b) donor andacceptor dyes (at positions D2 and A1, respectively, in Figure 1a). The 2Dplots are of FRET efficiency (E) or donor anisotropy (rD) versus donorlifetime (τD(A)). The grayscale indicates an increasing number of single-molecule bursts (from white to black). Also shown are the corresponding1D histograms. FRET efficiencies were measured from the raw green andred signals and corrected for background (1.54 kHz in green; 2.24 kHz inred), spectral crosstalk (3.5%), detection efficiencies (green/red = 0.3),and the fluorescence quantum yields (0.80 for the donor and 0.32 for theacceptor). The overlaid red line is the theoretical FRET relationship E = 1þ τD/τD(A) computed using τD = 4.1 ns. The overlaid blue line is thePerrin equation, rD = r0(1 þ τD(A)/FD), in which the mean rotationalcorrelation time (FD) was 0.35 ns and the fundamental anisotropy (r0) was0.375. The sample buffer contained 1 mM MgCl2.

Figure 1. (a) Schematic of the four-stranded fork (4SF) showing theDNA sequence and the positions of the donor (D) and acceptor (A) dyesfor single-molecule FRET measurements. A simplified representation isdepicted on the right. (b) Geometric model of a 4SF wherein the globalstructure is defined by the polar angles θ and φ and the angle of rotationaround the helical axis, ω.

Table 1. FRET Distances Calculated from the MFD Dataa

FRET pair

DA distance in

0 mM Mg2þ (Å)

DA distance in

1 mM Mg2þ (Å)b

D1A1 63 66

D1A2 61 62

D1A3 74 69

D2A1 66 69

D2A3 79 80

D3A1 66 65

D3A2 64 64

D3A3 68 68aThe distances were calculated from the donor lifetimes in the presenceof the acceptor by fitting the lifetime histogram of the FRET subpopula-tion in theMFD plots (e.g., Figure 2b) to a single Gaussian, as describedin the Supporting Information. The experiments used a buffer contain-ing 20 mM Tris, 15 mM NaCl, and 1 mM ascorbic acid at pH 7.5. Thestandard deviation was 2 Å for all of the distances except D1A3 at low[Mg2þ], where it was 3 Å. b In these experiments, the buffer alsocontained 1 mM MgCl2.

1190 dx.doi.org/10.1021/ja108626w |J. Am. Chem. Soc. 2011, 133, 1188–1191

Journal of the American Chemical Society COMMUNICATION

bending are small,13 the helical arms of the structures weremodeled and restrained as B-DNA. Three different sets of re-straints were tested: all base pairs restrained (0-free), one basepair per arm left unrestrained (1-free), and two base pairs per armunrestrained (2-free). The restraints of the bases included notonly base pairing but also base stacking. The distance root-mean-square deviation (DRMSD) between the FRET distances and thedye positions in the model measures how well the model fits themeasuredFRETdistances (FigureS2 in theSupporting Information).As described in the Supporting Information, the 1-free caseyielded the lowest DRMSD values while having a small variance(Figures S3-S5). The structures with the five lowest DRMSDvalues were superimposed and are shown in Figure 3 forsolutions containing either 0 mM Mg2þ (blue) or 1 mM Mg2þ

(orange). The structures for a particular sample are very similarto each other, illustrating that we have a well-defined globalstructure in each case. All of the structures are consistent with fullbase pairing. The 4SF adopts an open, planar conformation, andthere is no evidence of coaxial stacking between any of the helicalarms, in contrast to the Holliday junction.6

There is a clear difference between the structures in the presenceand absence of Mg2þ ions. The combination of small but repro-ducible distance changes results in a significant change in the fittedglobal structure, with symmetrical twisting around the axis definedby the parent duplex (the A2/D2 arm). There is also an asymme-trical folding whereby the D1/D3 arm is bent toward the parentduplex by ∼40� while the A1/A3 arm is bent by ∼20� in theopposite direction.

In summary, we have applied quantitative SM-FRET and MDsimulations to determine the global structure of a branched DNAmolecule.We have studied themolecule in solution, free from inter-ference from sample heterogeneity and surface effects, and havefound that it adopts an open, planar conformation without coaxialstacking of arms. There is evidence for ion-induced folding via inter-actions with Mg2þ ions. The largely symmetrical structure of the4SF found here in the presence of Mg2þ ions may dictate the effici-ency of recognition of such structures during processing of damagedreplication forks and recombination intermediates. We will explorefurther the dependence of sequence and local ionic environment onthe conformations of thesemolecules. These structures will serve asthe benchmark for our ongoing studies of forked DNA and theirinteractions with repair enzymes and accessory proteins.

’ASSOCIATED CONTENT

bS Supporting Information. Sample preparation, experi-mental and computational methods, Figures S1-S5, and PDBfiles. This material is available free of charge via the Internet athttp://pubs.acs.org.

’AUTHOR INFORMATION

Corresponding Authorsteven.magennis@manchester.ac.uk

’ACKNOWLEDGMENT

We thank the BBSRC (BB/G00269X/1) for support andClaus Seidel, Ralf K€uhnemuth, Suren Felekyan, Stefan Marawske,Alessandro Valeri, and Evangelos Sisamakis for assistance indeveloping our MFD setup. We thank Richard Henchman forhelpful discussions. S.W.M. acknowledges the award of an EPSRCAdvanced Research Fellowship (EP/D073154).

’REFERENCES

(1) (a) Atkinson, J.; McGlynn, P. Nucleic Acids Res. 2009, 37, 3475–3492. (b) Heller, R. C.; Marians, K. J. Nat. Rev. Mol. Cell Biol. 2006, 7,932–943.

(2) Aldaye, F. A.; Palmer, A. L.; Sleiman, H. F. Science 2008, 321,1795–1799.

(3) Gu, H. Z.; Chao, J.; Xiao, S. J.; Seeman, N. C. Nature 2010, 465,202–205.

(4) Yin, P.; Choi, H. M. T.; Calvert, C. R.; Pierce, N. A.Nature 2008,451, 318–322.

(5) Ortiz-Lombardía, M.; Gonz�alez, A.; Eritja, R.; Aymamí, J.;Azorín, F.; Coll, M. Nat. Struct. Biol. 1999, 6, 913–917.

(6) Lilley, D. M. J. Q. Rev. Biophys. 2000, 33, 109–159.(7) (a) Seeman, N. C.; Lukeman, P. S. Rep. Prog. Phys. 2005, 68,

237–270. (b) Seeman, N. C. In Algorithmic Bioprocesses; Condon, A.,Harel, D., Kok, J. N., Salomaa, A., Winfree, E., Eds.; Springer-Verlag:Berlin, 2009; pp 205-214.

(8) Lyubchenko, Y. L.; Shlyakhtenko, L. S. Methods 2009, 47, 206–213.

(9) (a) Weiss, S. Science 1999, 283, 1676–1683. (b) Roy, R.; Hohng,S.; Ha, T. Nat. Methods 2008, 5, 507–516.

(10) (a) McKinney, S. A.; D�eclais, A.-C.; Lilley, D. M. J.; Ha, T.Nat.Struct. Biol. 2003, 10, 93–97. (b) Joo, C.; McKinney, S. A.; Lilley,

Figure 3. Global structure of the 4SF derived from SM-FRET distance constraints andMD simulations. The five lowest-energy structures are shown for the4SF in a buffer containing either 0mMMgCl2 (blue) or 1mMMgCl2 (orange). The average positions of the helical axes of the duplex arms are represented bylines. The angles defining the global structure in 0mMMg2þ areω1=(-47.0( 0.5)�,ω2=(6.4( 0.3)�,φ1=(-18.1( 0.4)�,φ2=(25.6( 0.3)�, θ1=(38.8(0.2)�, and θ2 = (88.9( 0.2)�; in 1 mMMg2þ they areω1=(-10.9( 0.6)�,ω2=(21.3( 2.3)�, φ1=(-16.4( 0.9)�, φ2=(26.4( 2.6)�, θ1=(77.6( 0.3)�,and θ2 = (66.9 ( 0.6)�. See Figure 1b for the definitions of these angles; the subscripts 1 and 2 refer to arms D1/D3 and A1/A3, respectively.

1191 dx.doi.org/10.1021/ja108626w |J. Am. Chem. Soc. 2011, 133, 1188–1191

Journal of the American Chemical Society COMMUNICATION

D. M. J.; Ha, T. J. Mol. Biol. 2004, 341, 739–751. (c) Person, B.; Stein,I. H.; Steinhauer, C.; Vogelsang, J.; Tinnefeld, P. ChemPhysChem 2009,10, 1455–1460. (d) Ying, L. M.; Green, J. J.; Li, H. T.; Klenerman, D.;Balasubramanian, S. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 14629–14634.(11) Clegg, R. M. Methods Enzymol. 1992, 211, 353–388.(12) (a) Rothwell, P. J.; Berger, S.; Kensch, O.; Felekyan, S.;

Antonik, M.; W€ohrl, B. M.; Restle, T.; Goody, R. S.; Seidel, C. A. M.Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 1655–1660. (b) Gansen, A.;Valeri, A.; Hauger, F.; Felekyan, S.; Kalinin, S.; T�oth, K.; Langowski, J.;Seidel, C. A. M. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 15308–15313.(13) Wo�zniak, A. K.; Schr€oder, G. F.; Grubm€uller, H.; Seidel,

C. A. M.; Oesterhelt, F. Proc. Natl. Acad. Sci. U.S.A. 2008, 105,18337–18342.(14) Kalinin, S.; Sisamakis, E.; Magennis, S. W.; Felekyan, S.; Seidel,

C. A. M. J. Phys. Chem. B 2010, 114, 6197–6206.(15) Mathew-Fenn, R. S.; Das, R.; Harbury, P. A. B. Science 2008,

322, 446–449.

’NOTE ADDED AFTER ASAP PUBLICATION

Table 1, footnote a, was corrected January 11, 2011.